The Relative Industry Valuation Hypothesis of Cross-listing *

Transcription

1 The Relative Industry Valuation Hypothesis of Cross-listing * Kee-Hong Bae Schulich School of Business York University kbae@schulich.yorku.ca Yi Ding CUHK Business School The Chinese University of Hong Kong, Shenzhen yiding@cuhk.edu.cn Xiaoqiao Wang School of Business Nanjing University xiaoqiao.wang@nju.edu.cn This Version: May 2016 Keywords: Cross-listing; Relative industry valuation; Industry segmentation JEL Classification Code: F30, G32, G15. * We thank seminar participants at York University, Seoul National University, Korea University, Hong University of Science and Technology as well as conference participants at the Korean American Finance Association meeting. Comments from Andrew Karolyi and Sergei Sarkissian are particularly appreciated. All errors are our own.

2 The Relative Industry Valuation Hypothesis of Cross-listing Abstract We propose a new hypothesis of cross-listing the relative industry valuation hypothesis. We argue that for a firm considering cross-listing on the U.S. stock market, valuation discrepancy between the firm s home industry and the corresponding U.S. industry is an important factor for cross-listing decision. Consistent with the predictions of the hypothesis, we find that the extent of home industry undervaluation is positively related to the probability of cross-listing; significant cross-listing premiums are obtained for cross-listed firms from undervalued industries in the home markets; and higher valuation of these firms subsequent to listing leads to more equity capital raising activity.

3 1. Introduction Why do firms cross-list? Numerous theoretical and empirical studies of cross-listings have investigated firms motives for cross-listing and presented various hypotheses on this issue. 1 Nevertheless, there still appears to be significant uncertainty about the relative economic magnitudes of the valuation changes before and after the listing, how permanent or transitory they are, how quickly they disappear after initial listing, and which types of cross-listing firms experience the most dramatic changes (Karolyi, 2012, p. 527). We propose a new hypothesis of cross-listings relative industry valuation hypothesis, and provide a novel explanation on these questions. The relative industry valuation hypothesis predicts that for firms considering cross-listing their shares overseas, the difference between their industry valuation in the home market and the corresponding industry valuation in the target foreign market (i.e., the relative industry valuation) is an important factor in the cross-listing decision and in determining the valuation premium after listing. Some firms may have a low valuation because their industry valuation in the home market is generally depressed. Such firms have a greater incentive to cross-list their shares on the target stock market, particularly when the magnitude of relative industry undervaluation is large. Such firms will also likely realize a higher cross-listing premium once they are listed. We test the relative industry valuation hypothesis and find strong support for it. There are at least two reasons to suspect that the relative valuation between home country s industry and target country s industry is an important factor for a firm considering cross-listing. First, our measure 1 See survey articles by Karolyi (1998, 2006, 2012) for different hypotheses of cross-listing. 1

4 of relative industry valuation has the interpretation of industry segmentation as proposed by Bekaert, Harvey, Lundblad, and Siegel (2011). They propose the absolute differential between local and global valuation ratios as a model-free measure of market segmentation. 2 They find that while the level of country segmentation has decreased for many countries over time as a result of globalization, some industries still remain segmented. The level of industry segmentation has increased for nearly half of the industries in their sample. Consistent with the finding by Bekaert et al. (2011), Carrieri, Errunza, and Sarkissian (2004) argue that country-level integration does not preclude industry-level segmentation. To the extent that our measure of relative industry valuation captures the degree of industry segmentation, one should expect that firms from more segmented industries are more likely to cross-list and these firms realize higher valuation premiums, as cross-listing allows their integration into the global industry. In this sense, one can regard the relative industry valuation hypothesis as a variant of market segmentation hypothesis, or industry segmentation hypothesis (Stapleton and Subrahmanyam, 1977; Errunza and Losq, 1985; Eun and Janakiramanan, 1986; Alexander, Eun, and Janakiramanan, 1987). Second, the extant literature on initial public offering (IPO) shows that firms consider industry valuation as an important factor in their decision to go public (Ritter, 1984; Ritter, 1991; Lerner, 1994; Loughran and Ritter, 1995; Rajan and Servaes, 1997; Pagano, Panetta, and Zingales, 1998). Essentially, the cross-listing decision is a way of going public in a foreign stock market. In an increasingly globalized market, firms considering going public can choose to do so either in their home market or in the foreign 2 Our measure of relative industry valuation is different from the segmentation measure proposed in Bekaert et al. (2011) in several respects. We discuss the difference in Section 2. 2

5 market. For such firms, the valuations of industry peers from both home and foreign markets matter in their decision to go public. For firms considering cross-listing, the consideration of industry valuation is not restricted to the domestic market but rather is extended to cross-border markets. A high valuation of the firm s industry in the target market motivates the firm to cross-list, especially when the firm s home industry valuation is depressed. We focus on the U.S. stock market as a target market for firms to be cross-listed on as it is the largest recipient of cross-listed firms. Using a dataset of 2,891 cross-listed firms in the U.S. from 42 countries and covering 51 industries over the period 1982 to 2012, we find three sets of empirical results that together paint a coherent picture of the relative industry valuation effect on the cross-listing decision and valuation premium. First, we investigate whether higher relative industry valuation, measured as U.S. industry Tobin s q minus home industry Tobin s q, is associated with a higher probability of cross-listing. We find that most cross-listed firms are concentrated in industries that are relatively undervalued compared to the corresponding U.S. industries. When we apply the Cox proportional hazard model to quantify the relationship between the relative industry valuation and the cross-listing decision, we find that the probability of listing significantly increases when the U.S. industry Tobin s q is high and the home industry Tobin s q is low or when the relative industry valuation is high. Second, we investigate how relative industry valuation affects cross-listing premium. We compute the cross-listing premium as the average difference in Tobin s qs before and after listing. Our sample of cross-listed firms realize significantly positive listing premiums subsequent to listing. Tobin s q increases 3

6 by on average after cross-listing, which represents 9% of the average Tobin s q. When we partition the cross-listed firms by the median of their relative industry valuation before listing, we find a significantly positive cross-listing premium for the subsample with high relative industry valuation, whereas we find no cross-listing premium for the subsample with low relative industry valuation. For the subsample with high relative industry valuation, Tobin s q increases by an impressive magnitude of 0.238, which represents about 14% of average firm valuation. This finding is robust for both exchange listings and non-exchange listings, while the magnitude of the valuation premium is higher for exchange listings. We also examine whether the relative industry valuation effect on firm valuation are permanent or transitory and find that only firms from highly undervalued home industries enjoy permanent valuation gains through cross-listing. One major concern with the relative industry valuation hypothesis is that the positive association between relative industry valuation and the cross-listing premium is consistent with the bonding theory of cross-listing. Firms from countries with weak shareholder protection rights tend to have lower valuations (La Porta, Lopez-de-Silanes, Shleifer, and Vishny, 2002). These firms choose to bond themselves by cross-listing their stocks on better-governed stock markets, and they realize higher equilibrium valuations subsequent to listing as their governance risks are reduced (Stulz, 1999; Coffee, 1999). If the relative industry valuation effect on the valuation premium is simply a manifestation of the bonding effect, one should expect the relative industry valuation effect to be more pronounced for cross-listed firms from markets with poor investor protection. Our results indicate that the relative industry valuation effect remains the same regardless of the degree of investor protection, suggesting that our evidence is not due to bonding effect of cross-listing. We also find that firms from more developed markets tend to get higher valuation 4

7 benefits from cross-listing their shares when the relative industry valuation is large, suggesting that the relative industry valuation effect is distinct from the effects of traditional market segmentation hypothesis but rather is consistent with industry segmentation hypothesis. Firms with certain industry characteristics appear to realize more benefits of the relative industry valuation effect. The relative industry valuation effect on the valuation premium is likely more pronounced in industries where the industry factor in pricing equities is more important. We hypothesize that the industry factor is more important for firms in traded-goods industries than for firms in non-traded-goods industries (Griffin and Karolyi, 1998) and in industries with which U.S. investors are more familiar, and thus firms in these industries enjoy higher valuation premiums from relative industry valuation. We find strong support for this prediction. Finally, we investigate whether cross-listed firms issue more equity to utilize the benefits of higher valuation gains subsequent to listing. We find that both the number of firms raising equity and the dollar amount of equity raised increase significantly after cross-listing, but only for firms whose home industry valuation relative to the U.S. industry valuation is low. Our results are consistent with findings from the bonding literature (Reese and Weisbach, 2002; Doidge, Karolyi, and Stulz, 2009) that firms tend to issue more equity after cross-listing in the U.S. market. While their explanations of the increased capital-raising activities focus on the increased governance benefits, we find that firms may fuel their reserved capital needs via new equity issues after their undervaluation is removed by the cross-listing activity. Putting all the findings together, our study is consistent with the view of cross-listing in which undervaluation of the home industry relative to the U.S. industry motivates firms to cross-list in the U.S. 5

8 market, increases the firms valuations after cross-listing, and drives the firms to issue new equity to utilize the valuation gains. Our study relates and contributes to several stands of literature. First, our relative industry valuation hypothesis complements the market segmentation hypothesis of cross-listing. The traditional market segmentation hypothesis fails to explain many counterfactuals to the hypothesis. For instance, it fails to explain why the listing premium remains large even though investment barriers were eroding steadily over time; why price adjustments for firms from developed and emerging markets are so similar at cross-listing; why valuation changes vary so differently among firms listing from the same country; and why many firms that could qualify for a listing in the U.S. reject such option given large benefits of cross-listing (Karolyi, 2012). From the perspective of industry segmentation, our findings are able to explain these counterfactuals to the market segmentation hypothesis. Note that country-level integration does not preclude industry-level segmentation (Carrieri, Errunza, and Sarkissian, 2004). In fact, Bekaert et al. (2011) show that some industries still remain segmented, although country-level segmentation has significantly decreased over time. Thus, firms from segmented industries of developed markets may be equally better off as emerging market firms at the time of cross-listing. Valuation premium can be different among firms listing from the same domicile when they are from different industries with different degree of industry segmentation. Firms that could qualify for cross-listing may choose not to do so when their industry valuations are not different from target country s industry valuation, i.e., when their industries are not segmented. Second, our study is related to the debate on the relative importance of country factor versus industry factor in pricing equities in the international market. On the one hand, industry factor has become 6

9 increasingly important thanks to globalization (Roll, 1992; Baca, Garbe, and Weiss, 2000; Cavaglia, Brightman, and Aked, 2000; Ferreira and Gama, 2005; Campa and Fernandes, 2006; Baele and Inghelbrecht, 2009). On the other hand, despite globalization, the impact of financial globalization has been quite limited due to institutional frictions preventing ability of a country to take advantage of financial globalization (Stulz, 2005). As a result, country factor has continued to dominate industry factor in pricing international equities (Heston and Rouwenhorst, 1994; Griffin and Karolyi, 1998; Bekaert, Hodrick, and Zhang, 2009). Consistent with this literature, our evidence indicates that many industries remain segmented from global industry, and cross-listings continue to be important vehicles to fill the industry segmentation gap. Third, our study sheds light on the conflicting evidence on the existence of valuation premiums from cross-listing. Bonding hypothesis predicts a permanent valuation gain for cross-listed firms. Doidge, Karolyi, and Stulz (2004, 2009) document evidence of a significant cross-listing valuation premium for cross-listed firms over similar firms that do not cross-list on the U.S. markets. In contrast, time-series studies of valuation changes around cross-listing document a post-listing decline in valuations following cross-listing, suggesting that valuation gains from cross-listing are temporary (Gozzi, Levine, and Schmukler, 2008; Sarkissan and Schill, 2009, 2012). Our findings indicate that cross-listed firms from industries undervalued in their home markets relative to the U.S. market tend to enjoy permanent valuations gains, whereas cross-listing premiums tend to be temporary for those cross-listed firms from home industries with small relative undervaluation. 7

10 Finally, our study is related to the literature on equity market timing. Firms tend to go public or issue equity when stock market valuations are high (Loughran, Ritter, and Rydqvist, 1994; Pagano, Panetta, and Zingales, 1998; Jung, Kim, and Stulz, 1996; Hovakimian, Opler, and Titman, 2001; Graham and Harvey, 2001). International firms cross-list their stocks to benefit from the high valuation of the U.S. stock market, and these firms valuations quickly decline post-listing (Pagano, Roell, and Zechner, 2002; Gozzi, Levine, and Schmukler, 2008; Sarkissian and Schill, 2009, 2012). Consistent with the market timing hypothesis, we find that Tobin s q rises before cross-listing and then falls afterwards, but only for the cross-listed firms whose relative industry undervaluation is small or non-existent. Firms whose relative industry undervaluation is large maintain their valuation gains years after cross-listing. The rest of the paper proceeds as follows. In Section 2, we describe the data and empirical methodology. We investigate the impact of relative industry valuation on the cross-listing decision in Section 3 and on the cross-listing premium in Section 4. Section 5 examines capital raising activity after cross-listing and its relation to relative industry valuation. Section 6 concludes the paper. 2. Data Our sample selection starts with a list of countries for which the stock market capitalization of domestic stock exchanges is in excess of $100 billion as of We then include all firms in the Worldscope database from 1980 to 2012, which results in 39,979 firms from 43 countries. Of these firms, 4,733 are cross-listed on the U.S. stock exchanges. We obtain firms cross-listing information from Datastream, the Center for Research in Security Prices (CRSP), Citibank, the NYSE, Nasdaq, SEC Form 8

11 20-F filings, and Factiva searches. We manually cross-check and verify the information on a firm s listing activities across the various datasets. We keep track of both active and inactive issues and determine a firm s cross-listing status according to its status at the end of the calendar year. We use Tobin s q as our firm valuation measure. We compute Tobin s q as the market value of equity plus the book value of debt (computed as the book value of total assets minus the book value of equity) divided by the book value of total assets. All of the data used to compute Tobin s q are from Worldscope. Following the practice in the literature (e.g., Doidge, Karolyi, and Stulz, 2004, 2009; Gozzi, Levine, and Schmukler, 2008), we do not attempt to calculate the replacement cost of assets in the denominator and the market value of debt in the numerator since the required data are generally not available for most of the sample firms. There are several advantages of using Tobin s q as the firm valuation measure. First, it is a forwardlooking and risk-adjusted measure. Second, it is comparable across international firms, which is particularly important in our context as our main measure of relative industry valuation is based on the comparison of the home industry Tobin s q to that of the U.S. industry. Finally, it is well grounded in economic theory as a measure of firm value. To reduce the impact of outliers on the test results, we winsorize the Tobin s q at the 1% level on both tails. We classify firms according to two-digit Standard Industrial Classification (SIC) codes, which have 62 different industry classifications. On the one hand, using three-digit SIC or other narrower industry classifications significantly reduces the number of observations in each industry and may increase the estimation bias of industry-level valuations for countries with a small group of firms. On the other hand, 9

12 using broader industry classifications may not provide enough cross-sectional variation across industries to distinguish between the effect of country variation and that of industry variation (Griffin and Karolyi, 1998). We also perform empirical analyses using the Industry Classification Benchmark (ICB) and find the key results unchanged. We compute U.S. industry q and home industry q as the average q of all firms within the firm s industry in the U.S. and the home country where the firm is domiciled, respectively. We exclude foreign listings to avoid double counting in computing industry q. We also exclude firms with total assets less than $10 million (in 2010 dollars) to make firms across countries more comparable. 3 We winsorize firm Tobin s q at the 1% level on both tails before calculating the industry Tobin s q to reduce the impact of outliers. In computing industry Tobin s q, we require that each industry in a country have at least five firms with available data to compute Tobin s q. As a robustness check, we also compute and conduct empirical analyses using the asset-weighted average Tobin s q of all firms in the industry and find the results unchanged. We compute relative industry q as the difference between U.S. industry q and home industry q. Higher relative industry valuation means that U.S. industry q is higher than home industry q, which suggests that firms in the home industry are relatively undervalued compared to firms in the same industry in the U.S. Bekaert et al. (2011) propose the absolute difference between local and global earnings yield (inverse of price to earnings ratio) as a model-free measure of segmentation. While similar in spirit, our measure of relative industry valuation is different from their segmentation measure in several respects. First, we use the signed difference between home industry valuation and U.S. industry valuation, while Bekaert 3 We also perform an analysis using all firms with assets of $100 million or more. The key results are not affected. 10

13 et al. (2011) use absolute difference. Second, we use U.S. industry valuation as a benchmark whereas they use global industry valuation as a benchmark. Third, we use Tobin s q as a measure of valuation, while they use earnings yield. While not reported, when we compute the degree of segmentation as absolute difference between home industry Tobin s q and global industry Tobin s q, we are able to confirm the main findings of Bekaert et al. (2011) that the level of market segmentation has decreased significantly for most markets, but the level of industry segmentation still remains for many industries. We compute several commonly used determinants of Tobin s q such as total assets, sales growth, return on assets (ROA), and global industry q. Total assets (in millions of U.S. dollars) are adjusted for inflation computed from local consumer price index (CPI) changes from the International Monetary Fund estimates. Sales growth is measured as the two-year geometric average of annual inflation-adjusted growth in sales 4 and is winsorized at the 1% level on both tails. ROA is calculated as earnings before interest and taxes scaled by total assets. Global industry q is the average Tobin s q of the industry to which a firm belongs and is calculated using all firms in the same industry. We exclude firms in the financial industry because they are highly leveraged and heavily regulated in the U.S. market and are likely valued differently from nonfinancial firms. After removing firms with missing data, our final sample includes 28,453 international firms with 256,062 firm-year observations. Among these firms, there are 2,891 cross-listed firms from 42 countries covering 51 industries during the period 1982 to 2012, resulting in 29,019 firm-year observations. 5 4 If the two-year growth rate is not available, one-year growth in sales is used. 5 The accounting data of cross-listed firms were generally not available in Worldscope until Since we require firms to have at least two years of sales data to calculate sales growth, firms that were cross-listed prior to 1981 are not included in the sample. 11

14 Table 1 shows the distribution of 2,891 cross-listed firms in the U.S. by industry and home country. The last row and column of the table report the total number of listings from each industry and each home country, respectively. The five largest suppliers of listings are Canada (807 listings), the United Kingdom (343), Australia (326), Hong Kong (261), and Japan (220). Our sample distribution is similar to that in Doidge, Karolyi, and Stulz (2004) and Gozzi, Levine, and Schumukler (2008). The five largest industries of listings are Oil & Gas Extraction (379 listings), Metal Mining (355), Business Services (238), Chemical & Allied Products (196), and Electronic & Other Electric Equipment (190). The mining industry is dominated by Australian and Canadian listings, whereas the listings for the electronic industry are primarily from Asian countries. Table 2 lists the name of industries, the number of cross-listings in each industry, and the mean values of U.S. industry q, home industry q and relative industry q during the entire sample period. Several observations emerge. First, more firms are cross-listed from industries where the relative q is high. Note that Table 2 lists the industries by the magnitude of relative industry q. The five industries with highest relative industry q are Chemical & Allied Products, Nonmetallic Minerals, Metal Mining, Transportation Services, and Electronic & Other Electric Equipment. It is notable that Chemical & Allied Products, Metal Mining, and Electronic & Other Electric Equipment are among the five largest cross-listing industries. In fact, the top five industries in the magnitude of relative industry q represent 26.7% (776) of all cross-listed firms, whereas the bottom five industries in relative industry q account for only 3.9% (114). Second, although both U.S. and home industry qs vary across industries, U.S. industries are generally valued higher than cross-listed firms home industries. The average U.S. industry q is 1.92, whereas the average home 12

15 industry q is 1.55, being undervalued by about 20% relative to the U.S. industry q. In 45 of the 51 industries, the U.S. industry q is higher than the home industry q. 3. Relative industry valuation and the cross-listing decision In this section, we investigate a firm s cross-listing decision using a Cox (1972) proportional hazard model. The model has been extensively used in studies of the cross-listing decision and is particularly suited to the prediction of discrete events in a panel setting (Pagano, Roell, and Zechner, 2002; Claessens, Klingebiel, and Schmukler, 2003; Doidge, Karolyi, and Stulz, 2009). We model the probability of crosslisting in year t+1, given that the firm has not yet listed, as a function of firm and country characteristics variables in year t. These variables include (i) firm size measured by the logarithm of the firm s total assets (in millions of U.S. dollars), (ii) firm leverage, (iii) growth rate of sales over the previous two years, (iv) ROA, (v) real GDP growth rate, (vi) logarithm of the ratio of stock market capitalization over GDP, and (vii) global industry q of each firm s industry. We report the coefficients on the explanatory variables in exponential form, which can be interpreted as the effect of a unit change in the explanatory variable on the hazard ratio: for example, an exponential coefficient of 1.1 implies that a unit change in the explanatory variable increases the hazard ratio (i.e., the probability of a cross-listing event) by 10%. We adjust the standard errors for clustering at the country-industry-year level because our main variable of interest, relative industry q, changes at the country-industry-year level. Table 3 presents the results. Columns (1) to (3) present how firm q, home industry q, and U.S. industry q, respectively, affect the cross-listing decision of firms. Column (1) presents the result using firm 13

16 Tobin s q and other controls as explanatory variables. The hazard ratio on firm Tobin s q is positive and significantly greater than 1, suggesting that high-valued firms are more likely to cross-list. In column (2), we add home industry q as an explanatory variable in addition to control variables, including firm Tobin s q. Surprisingly, the hazard ratio on the home industry q is and significantly different from one (zstatistic = -4.74). This result suggests that firms with a high home industry valuation are less likely to crosslist. In column (3), we replace home industry q with the U.S. industry q. The hazard ratio on the U.S. industry q is and significantly different from one (z-statistic = 3.74). Firms are more likely to crosslist when their industry valuation in the U.S. market is high. In column (4), we add both home industry q and U.S. industry q at the same time and confirm that firms with a low home industry valuation and a high U.S. industry valuation are more likely to cross-list. In column (5), we report the results using relative q, calculated as the difference between U.S. and home industry q, as the proxy for relative industry valuation or home industry undervaluation. The result shows that the hazard ratio for the relative q is and statistically significantly different from one (zstatistic = 5.42), suggesting that a one-unit increase in relative q increases the hazard ratio by 33.3%. One may argue that the impact of relative industry valuation on the cross-listing decision is mainly driven by the U.S. industry valuation. Put differently, it is not relative industry valuation between the U.S. and home country but high U.S. market valuation that motivates a firm to cross-list. We argue that this is not the case for two reasons. First, we show that home industry q affects the cross-listing decision even when we control for U.S. industry q. Further, a high home industry valuation discourages cross-listing. Second, in unreported tests, we find that the effect of home industry q or U.S. industry q on cross-listing disappears when we 14

17 include the relative q in the regression. This result suggests that firms are more likely to cross-list when their home industries are undervalued relative to the corresponding U.S. industries and that the relative industry valuation effect is not entirely driven by high U.S. industry valuation alone. 4. Relative industry valuation and the cross-listing premium In this section, we examine how relative industry valuation affects the cross-listing premium, measured as the change in the Tobin s q of cross-listed firms before and after cross-listing. We first document evidence that the cross-listing decision is associated with a positive premium and that this premium is more evident for cross-listed firms from undervalued home industries. We then show that the effect of relative industry valuation on the valuation premium is not driven by the two leading hypotheses in the cross-listing literature, market segmentation and bonding, but complements both hypotheses. Finally, we conduct several robustness tests to further support the relative industry valuation effect The effect of relative industry valuation on the cross-listing premium We measure cross-listing premium as the change in Tobin s q of cross-listed firms before and after cross-listing. We use the pre-listing relative q, calculated as the difference between U.S. industry q and home industry q one year before cross-listing, to measure the extent of pre-listing home industry undervaluation and require sample firms to have non-missing data on this variable. These restrictions reduce our sample to 28,925 firm-year observations for 2,601 cross-listed firms. To get an overall picture of the relative industry valuation effect on the cross-listing premium, we divide the cross-listed firms into two groups according to the median value of the pre-listing relative q and 15

18 plot the evolution of Tobin s q over the three-year period before and after cross-listing for these two groups. We define year t as the year of cross-listing on the U.S. market. Year t-k and year t+k are k year before and after cross-listing, respectively. We compute the average Tobin s q for cross-listed firms each year relative to the Tobin s q in year t-3 to facilitate the comparison of valuation changes between the two groups of cross-listed firms. Figure 1 presents the relative valuation changes. For cross-listed firms with low relative industry valuation, Tobin s q peaks at t-1 and quickly declines in subsequent years. By year t+3, these firms end up with even lower valuation levels than their valuation levels three years before cross-listing. This finding is consistent with Gozzi, Levine, and Schmukler (2008) and Sarkissan and Schill (2009, 2012), who show that cross-listing premiums do not last. By contrast, for firms with high relative industry valuation, their Tobin s q peaks at the year of cross-listing. Note that we measure Tobin s q in the year of cross-listing at the end of calendar year, so that an increase in q in year t from t-1 suggests that a firm s value increases immediately after cross-listing: that is, the relative undervaluation is adjusted right after listing. We also note that the post-listing valuation premium tends to persist by year t+2. Table 4 presents summary statistics on the average Tobin s q of cross-listed firms during the three years before and after cross-listing. It also presents the average Tobin s q for the subsamples partitioned by the level of relative industry valuation one year before cross-listing. We divide cross-listed firms into two groups according to the median value of the pre-listing home industry undervaluation, and for each group, we calculate the average firm q for the three years before and after cross-listing. Columns (1) to (3) provide summary statistics for all types of cross-listings in the U.S., while columns (4) to (6) and (7) to (9) 16

19 respectively provide the statistics for exchange listings and non-exchange listings. The results indicate that after listing, the average Tobin s q slightly increases to from before listing. However, the increase is not significant. When we partition the cross-listed firms by the level of relative industry valuation, we find strikingly different results. For firms with high relative industry valuation, the average firm q significantly increases by 0.16 after cross-listing, whereas it significantly decreases by 0.09 for firms with low relative industry valuation. We further examine the valuation gains for exchange listings in columns (4) to (6) and non-exchange listings in columns (7) to (9) and find that valuation gains are concentrated in firms with high relative industry valuation. We now formally test whether cross-listed firms have valuation gains surrounding cross-listing and whether the extent of relative industry valuation affects the valuation premium using the regression framework. Table 5 presents the results. We use Tobin s q as the dependent variable for a panel of crosslisted firms across the period 1982 to The main independent variable of interest is the cross-listing dummy that equals one on and after the year when a firm cross-lists in the U.S. and zero otherwise. Delisted firms after cross-listing are present in the sample until the year of delisting. The coefficient on the cross-listing dummy measures the average increase in firm valuation after cross-listing, the cross-listing premium. Column (1) shows the results of regressing the Tobin s q of all firms on the cross-listing dummy and other control variables. The cross-listing dummy has a positive coefficient estimate of with a t- statistic of 4.46, which indicates a positive and significant cross-listing premium for the sample firms. This valuation premium represents about 9% of the average firm q. The coefficient estimates on the control variables are consistent with expectations. Larger and high-leveraged firms tend to have lower valuations, 17

20 while high-growth, profitable firms tend to command higher valuations. Not surprisingly, global industry q is positively related to firm Tobin s q. In columns (2) and (3), we divide the sample into two subsamples according to the median of the pre-listing relative q and repeat column (1). The results are striking. The relation between the cross-listing dummy and firm Tobin s q is significantly positive for firms with high relative industry valuation (0.238 with a t-statistic of 4.92), whereas it is insignificant for firms with low relative industry valuation. The difference in the coefficient estimates on the cross-listing dummy between the two subsamples is statistically significant at the 1% level. Our results on the cross-listing premium indicate that valuation gains from cross-listing exist only for the firms from undervalued industries in the home markets. In columns (4) to (6), we repeat columns (1) to (3) for the sample of exchange listings 6 and find consistent results: that is, firms from undervalued industries in the home markets enjoy a large and significant cross-listing premium, whereas those from industries less or not undervalued at home do not obtain a cross-listing premium. We examine the cross-listing premiums of non-exchange listings in columns (7) to (9). Column (7) shows that the premium for non-exchange listings is significantly positive but its magnitude is lower than that of exchange listings. This results is consistent with Doidge, Karolyi, and Stulz (2004), who argue that OTC markets require only minimal SEC disclosure and no GAAP compliance. These firms are also exempt from filing Form 20-F under Rule 12g3-2(b), which allows home country accounting statements with an adequate English translation, if necessary. Other non-exchange listings similarly meet 6 We classify firms that were listed over the counter and on the exchange at different points during this 30-year period as exchange listings. 18

21 less regulatory requirements than exchange listings. Therefore, non-exchange listings experience less bonding benefits. Consistent with the results using exchange listings, when we divide the sample by the magnitude of home industry undervaluation, we find that there is a significantly positive cross-listing premium only for firms with high relative industry valuation. The difference between the two subsamples is statistically significant at the 5% level. Although non-exchange listed firms appear to obtain less bonding benefits, they still enjoy a cross-listing premium when their home industries are undervalued. Finally, while not reported, to further examine whether our findings are merely driven by U.S. industry q or home industry q alone, we replicate the analyses by partitioning the sample by the median of U.S. industry q or home industry q. We find no significant difference in the magnitude of cross-listing premiums between the subsamples, suggesting that our results are not driven by U.S industry q or home industry q alone. In Table 6, we examine how the valuation premium evolves over time after cross-listing. Specifically, we break down the cross-listing dummy in Table 5 into dummy variables that take the value of one in each year on and after cross-listing. The dummy for T+k takes the value of one in the k year after cross-listing and zero otherwise, where k ranges from 0 to 3. The dummy for T+4 or greater takes the value of one in the four years or more after cross-listing and zero otherwise. Columns (1) to (3) show the evolution of valuation for all types of listings. We find that all firms have a valuation gain in the year of cross-listing compared to the time before cross-listing. Moreover, the valuation gain remains significant and positive for every year after cross-listing. When we partition the sample into high and low relative industry valuation subsamples, we find there is a significant increase in 19

22 firm s valuation for both subsamples in the year of cross-listing. However, for firms with low relative industry valuation, Tobin s q falls quickly in the years after cross-listing. One year after cross-listing, Tobin s q loses nearly all its valuation gain from the previous year and is no longer significantly higher than its value before cross-listing, and it never rises again in the following years. For firms with high relative industry valuation, their valuation gains remain positive for each and every year after cross-listing. Interestingly, the coefficient estimate on the dummy for T+4 or greater, which represents all the years since the fourth year following cross-listing, is strongly positive, suggesting a permanent valuation gain for cross-listings. In columns (4) to (6) and (7) to (9), we further examine the time-series evidence of the crosslisting premium for exchange and non-exchange listings separately. We find strikingly similar evidence for both exchange and non-exchange listings. In sum, the results in Table 5 and Table 6 provide strong evidence that there are significant valuation gains associated with cross-listing and that these valuation gains exist for firms from industries that are undervalued in their home markets relative to the U.S. market. Doidge, Karolyi, and Stulz (2004, 2009) document evidence of a significant cross-listing valuation premium of cross-listed firms compared to home-market firms that have similar firm characteristics but do not cross-list on U.S. markets. Their evidence on the cross-listing premium is cross-sectional by nature and does not consider the evolution of valuation changes over the longer time horizon. In contrast, time-series studies of valuation changes around cross-listing mostly document a post-listing decline in valuation following cross-listing, suggesting that valuation gains from cross-listing are temporary (Gozzi, Levine, 20

23 and Schmukler, 2008; Sarkissan and Schill, 2012). 7 Our findings may shed light on the conflicting evidence in the previous literature on the cross-listing premium. Our results indicate that cross-listed firms from relatively undervalued industries in home markets tend to enjoy permanent valuation gains, whereas the cross-listing premium tends to be temporary for those cross-listed firms that suffer less from home industry undervaluation Is the relative industry valuation effect on the cross-listing premium driven by market segmentation and bonding? In this subsection, we examine whether the relative industry valuation effect on the cross-listing premium can be explained by the two leading theories in the cross-listing literature: market segmentation and bonding. Market segmentation theory predicts that firms in segmented markets will have depressed firm valuation due to investment barriers and that cross-listed firms enjoy the benefits of tapping into developed capital markets, leading to higher valuation. Thus, firms from more undervalued industries in home markets realize higher valuation premiums because they are more segmented from the global capital market, leading to greater benefits of integration. We argue that if the relative industry valuation effect on the cross-listing premium is simply a manifestation of the market segmentation effect, one should expect the relative industry valuation effect to be more pronounced for cross-listed firms from less developed, segmented markets. 7 A notable exception is King and Siegel (2009), who find that using a Canadian sample of cross-listed firms, crosslisted firms with a single class of shares enjoy a permanent increase in valuation if they maintain investor recognition over time. 21

24 The bonding theory predicts that firms from countries with weak shareholder protection suffer from agency conflicts; such firms can choose to bond themselves to markets with stronger legal and financial institutions by cross-listing their stocks on better-governed stock markets. These firms then enjoy a higher equilibrium valuation as their governance risks are reduced. La Porta, Lopez-de-Silanes, Shleifer, and Vishny (2002) find that firms from countries with better shareholder protection laws have a higher valuation than comparable firms from countries with weaker shareholder protection laws. Thus, if firms in relatively undervalued industries in their home markets are in fact those from markets with a weak governance structure, then these firms will enjoy a valuation premium when cross-listed not because of the relative industry valuation effect but because of the bonding effect. We argue that if the relative industry valuation effect on the valuation premium is simply a manifestation of the bonding effect, one should expect the relative industry valuation effect to be more pronounced for cross-listed firms from markets with poor investor protection. To examine the concern that our results are due to the market segmentation effect or the bonding effect, we repeat the regression analyses in Table 5 for the subsamples partitioned by the degree of market development/segmentation and investor protection; the results are presented in Table 7. We measure the degree of market development/segmentation by the ratio of stock market capitalization over GDP and divide the sample firms according to its median value. A higher value of this ratio represents a higher degree of market development or a lower degree of market segmentation. We use the country s legal origin as the proxy for country-level shareholder protection. We consider a firm as having strong shareholder protection laws when it is from a common law country. 22

25 The first four columns (1) to (4) of Table 7 present the relation between relative industry valuation and the cross-listing premium for the subsamples partitioned by the degree of market development: columns (1) and (2) for less developed markets and columns (3) and (4) for more developed markets, respectively. We find that the valuation premium is higher for firms with high relative industry valuation regardless of the extent of market development. However, contrary to the prediction of the market segmentation theory, we find that the cross-listing premium is much higher for firms from more developed markets. In column (2), the cross-listing premium for firms from less developed markets is with a t-statistic of 3.0 when the relative industry valuation is high. In column (4), the corresponding figure for firms from more developed markets is with a t-statistic of This finding suggests that firms from more developed markets enjoy a significant valuation gain when their relative industry valuation is high. Thus, the relative industry valuation effect on the cross-listing premium cannot be explained by the market segmentation predictions. For robustness, we use an alternative proxy for the extent of market development in unreported regressions. We partition our sample into firms from emerging and developed markets on the basis of the Morgan Stanley Capital International (MSCI) classification and repeat the regressions in columns (1) to (4). Consistent with the results using stock market capitalization as a measure of stock market development, we find that when the magnitude of relative industry valuation is high, firms from developed markets realize higher valuation premiums than those from emerging markets. 23

26 The last four columns (5) to (8) of Table 7 present the relation between relative industry valuation and listing premium for the subsamples partitioned by the extent of shareholder protection. 8 We find significant (insignificant) cross-listing premiums when relative industry valuation is high (low) regardless of a firm s legal origin. The comparison of estimates on the cross-listing dummy in columns (6) and (8) also indicates that the magnitude of cross-listing premiums is remarkably similar in both civil law and common law countries, suggesting that legal investor protection is not the main driver of the relative industry valuation effect. One should see a more pronounced effect of relative industry valuation for firms in civil law countries if the bonding effect mainly drives our results. While the relative industry valuation effect is distinct from the traditional market segmentation and bonding hypotheses, it complements the two hypotheses. First, the discrepancy in industry valuations across different countries could be induced by market segmentation or governance reasons in the first place. Using industry-portfolio valuation-based measure of market segmentation, Bekaert et al. (2011) show that factors such as market openness, political risk profile, financial development and business cycle together contribute to segmentation across country and industry. A large literature documents the linkage between corporate governance and stock market valuation to explain the cross-country valuation discrepancies (La Porta, Lopez-de-Silanes, Shleifer, and Vishny, 2002; Shleifer and Wolfenzon, 2002; Hail and Leuz, 2006; Daske, Hail, Leuz, and Verdi, 2008). 8 When we use the anti-self-dealing index developed by Djankov, La Porta, Lopez-de-Silanes, and Shleifer (2008) to measure investor rights protection, we find similar results. 24

27 Second, although our finding is contrary to the prediction of market segmentation hypothesis on cross-listing that firms from more segmented markets should enjoy higher cross-listing premiums, our results are not inconsistent with the process of market integration. To the extent that the relative industry valuation captures the extent of industry segmentation, firms from developed markets may be better able to utilize the industry segmentation alongside globalization, so that the relative industry valuation effect on cross-listing premium is more pronounced for these firms. We conjecture that the magnitude of cross-listing premium is determined not only by the ex-ante valuation discrepancy caused by segmentation or governance reasons, but also by the ex-post capability to adjust the valuation discrepancy. This may explain why firms from developed markets in the end enjoy higher listing premiums from the relative industry valuation effect. Similarly, while firms from countries with better governance may have lower ex-ante valuation discrepancy at industry level, they may still capture high cross-listing premium because they are better able to adapt to the U.S. market and remove the relative industry undervaluation Relative industry valuation effect on the cross-listing premium and industry characteristics In this subsection, we provide further evidence on the effect of relative industry valuation on the cross-listing premium and examine how industry characteristics interact with the relative industry valuation effect. We hypothesize that the relative industry valuation effect is more pronounced in industries where the industry factor in pricing equities is more important. Griffin and Karolyi (1998) argue that for industries that do not produce goods that are traded internationally ( nontraded-goods industries ), country factors explain a relatively larger proportion of the variation in stock returns. For industries that produce goods that 25